Basics of AI
Artificial Intelligence, Machine Learning and Deep Learning are all buzz words that seem to be everywhere these days. They’re in your smartphone, your smartwatch, your smart speakers and smart fridge. But what’s the difference between them, and are they r
Artificial Intelligence is a general term for any form of computer thinking. It can refer to anything from the AI opponent in a game of Starcraft, to a voice-recognition system like Siri or Google Now interpreting and responding to speech.
Furthermore, the technology can broadly be categorized into two groups: Narrow AI and Artificial General Intelligence (AGI).
Google’s DeepMind AlphaGo AI is an example of narrow AI: one that is skilled at a specific task. In contrast, an AGI is a general purpose AI that, in theory can apply its thinking to any task. A true AGI is much harder to create and while there have been many attempts to make one, so far, no AI can be classified as an AGI.
Machine learning is a subset of AI. It refers specifically to software designed to detect patterns and observe outcomes, then use that analysis to adjust its own behavior. Machine learning doesn’t actually require intelligent thinking in the way we perceive it, it simply requires really good pattern matching and the ability to apply those patterns to its behavior.
IBM’s Deep Blue and DeepMind’s Alpha Go are both game-playing Narrow AIs, but only Alpha Go uses Machine Learning. Deep Blue uses rule-based programming, so any changes in its behavior relies on changes in its core programming. On the other hand, Alpha Go was able to beat Go world champion Ke Jie by analyzing expert-level Go matches and applying those strategies.
Deep Learning is a further subset of Machine Learning that uses algorithms inspired by the structure of the human brain called artificial neural networks to solve problems. In Deep Learning, the AI have multiple layers that handle different specific tasks. And even something as simple as identifying a stop sign will have different layers to analyze shape, color, patterns, and text. As such, Deep Learning requires massive datasets to work. For example, Google’s self-driving cars are trained to recognize obstacles and react to them appropriately. But due to the infinite number of variables (other cars, pedestrians, weather and road conditions etc.) involved, Google requires a massive amount of data to analyze.